Vous êtes sur la page 1sur 7

Intelligence 44 (2014) 142–148

Contents lists available at ScienceDirect

Intelligence

Verbal intelligence is correlated with socially and economically


liberal beliefs
Noah Carl ⁎
Nuffield College, New Road, Oxford, OX11NF, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: Research has consistently shown that intelligence is positively correlated with socially liberal
Received 1 December 2013 beliefs and negatively correlated with religious beliefs. This should lead one to expect that
Received in revised form 14 March 2014 Republicans are less intelligent than Democrats. However, I find that individuals who identify
Accepted 15 March 2014 as Republican have slightly higher verbal intelligence than those who identify as Democrat
Available online xxxx
(2–5 IQ points), and that individuals who supported the Republican Party in elections have
slightly higher verbal intelligence than those who supported the Democratic Party (2 IQ points). I
Keywords: reconcile these findings with the previous literature by showing that verbal intelligence is
Intelligence correlated with both socially and economically liberal beliefs (β = .10–.32). My findings suggest
Democrats
that higher intelligence among classically liberal Republicans compensates for lower intelligence
Republicans
among socially conservative Republicans.
Socially conservative
Classically liberal © 2014 Elsevier Inc. All rights reserved.

1. Introduction hypothesis, Republicans are less likely to believe in widely


accepted scientific ideas such as climate change and the theory
Over the last few years, scholarly interest in the relationship of evolution (Kohut, Doherty, & Dimmock, 2009). Indeed,
between intelligence and political beliefs has grown consider- Mooney (2005) argues that, over the last couple of decades,
ably. A consistent finding is that people with higher intelli- members of the Republican Party have attempted to system-
gence tend to be more socially liberal (Deary, Batty, & Gale, atically undermine certain fields of scientific research. And in
2008a, 2008b; Stankov, 2009; Kanazawa, 2010; Schoon, Cheng, his latest book, Mooney (2012, pp. 59–126) contends that
Gale, Batty, & Deary, 2010; Heaven, Ciarrochi, & Leeson, 2011; Republicans' denial of science stems not only from perceived
Hodson & Busseri, 2012). Another consistent finding is that political advantage, but from psychological traits that incline
people with higher intelligence tend to be less religious (Bell, Republicans to prize certainty above all else.
2002; Lynn, Harvey & Nyborg, 2009; Nyborg, 2009; Ganzach, However, there is evidence pointing in the other direction.
Ellis, & Gotlibovski, 2013; Zuckerman, Silberman, & Hall, 2013). To begin with, education is correlated with the tendency to
Given that Republicans tend to be both more religious and think like an economist, which could be considered a centre-
more socially conservative than Democrats (Newport, 2007; right characteristic (Caplan, 2001; Caplan, 2007, pp. 50–93;
Saad, 2012), these two findings should lead one to expect Caplan & Miller, 2012). More importantly, intelligence itself
that Republicans have lower intelligence. Consistent with this is correlated with the tendency to think like an economist,
at least in the United States (Caplan & Miller, 2010). For
example, Americans with higher intelligence are less likely
Abbreviations: OLS, Ordinary Least Squares; PCA, Principal Components
Analysis.
to agree with statements such as “it is the government's
⁎ Tel.: +44 7791259551. responsibility to provide a job for everyone who wants one”,
E-mail address: noah.carl@nuffield.ox.ac.uk. and “corporations should pay more of their profits to workers

http://dx.doi.org/10.1016/j.intell.2014.03.005
0160-2896/© 2014 Elsevier Inc. All rights reserved.
N. Carl / Intelligence 44 (2014) 142–148 143

and less to shareholders”. They are also less likely to agree with that individuals with higher IQs have larger vocabularies
the statement, “it is the government's responsibility to reduce (Jensen, 2001). Vocabulary tests load more strongly onto the
the differences in income between people with high incomes crystallized factor of intelligence than onto the fluid factor,
and those with low incomes” (Kanazawa, 2010). Furthermore, so the test included in the GSS is most appropriately
Rindermann, Flores-Mendoza, and Woodley (2012) analyzed described as a measure of verbal intelligence, rather than
data from Brazil, and found that mean IQ was highest among problem-solving ability (Cattell, 1963; Horn & Cattell, 1966).
individuals who described themselves as centre-right. In For a longer discussion of the measure's validity, see Caplan and
addition, there is evidence that libertarians, who are more Miller (2010). Prior to analysis, I transform the measure so that
likely to vote for the Republican Party (Kirby & Boaz, 2010), it has a mean of 100 and a standard deviation of 15, which is the
have higher intelligence than both conservatives and progres- convention for normalizing IQ scores.
sives (Kemmelmeier, 2008; Iyer, Koleva, Graham, Ditto, & In the GSS, party identity is assessed with the question, “Do
Haidt, 2012). Finally, Republicans have better objective political you think of yourself as a Republican, Democrat, Independent,
knowledge than Democrats (Kohut, Doherty, Dimmock, & or what?” (Smith et al., 2012). There are eight response
Keeter, 2012). And just like conservatives, progressives are categories: “strong Democrat”, “not strong Democrat”, “Inde-
prone to logical fallacies and unscientific thinking (Berezow & pendent, near Democrat”, “Independent”, “Independent, near
Campbell, 2012). Republican”, “not strong Republican”, “strong Republican”, and
Pinker (2011, pp. 662–664), drawing on some of the “other”. I create three binary variables, corresponding to three
evidence outlined above, argues that intelligence is actually alternative definitions of party identity. The first variable,
correlated with classically liberal beliefs. According to McLean which uses a narrow definition of party identity, takes the
and McMillan (2009, pp. 306–308), classical liberalism is “the value ‘1’ if a respondent answered “strong Republican” and
belief that it is the aim of politics to preserve individual rights takes the value ‘0’ if he answered “strong Democrat”. The
and maximise freedom of choice” (see also Miller, 2003, second, which uses an intermediate definition, takes the value
pp. 55–73). Classical liberals define ‘liberty’ in the negative ‘1’ if a respondent answered “strong Republican” or “not strong
sense, as freedom from coercion and interference (Berlin, Republican” and takes the value ‘0’ if he answered “strong
1969, pp. 123–4). They hold both socially and economically Democrat” or “not strong Democrat”. The third, which uses a
liberal beliefs (Friedman, 1962, pp. 5–6). Socially liberal beliefs broad definition, takes the value ‘1’ if a respondent answered
are predicated on the idea that an individual should be free to “strong Republican”, “not strong Republican” or “Indepen-
pursue his own values and make his own lifestyle choices. dent, near Republican” and takes the value ‘0’ if he answered
Economically liberal beliefs are predicated on the idea that an “strong Democrat”, “not strong Democrat” or “Independent,
individual should be free to engage in voluntary transactions near Democrat”.
with others and to enjoy the fruits of her labour. Pinker's For each presidential election that took place between
(2011) hypothesis predicts that intelligence should be associ- 1968 and 2008, the GSS contains at least one wave in which
ated with economically liberal beliefs, as well as socially liberal respondents were asked how they voted in that election or
beliefs. how they would have voted if they did not (Smith et al., 2012).
For example, respondents interviewed in 1987, 1988 and 1989
2. Method were asked how they voted or would have voted in the 1984
election, while those interviewed in 1989, 1990, 1991 and 1993
2.1. Data were asked how they voted or would have voted in the 1988
election. Notice that respondents interviewed in 1989 were
I analyze data from the General Social Survey (GSS), a asked about the 1984 election, as well as the 1988 election. I
public-opinion survey that has been administered to a create two binary variables, corresponding to those who voted
nationally representative sample of American adults every
1–2 years since 1972. The GSS contains questions on respon-
dents' socio-economic characteristics, behaviours, and social
attitudes. It has been used by numerous previous studies to
examine intelligence (e.g., Kanazawa, 2010; Caplan & Miller,
2010; Carl & Billari, 2014). Each wave of the GSS provides data Table 1
on a cross-section of the U.S. population in a particular year. Difference in mean verbal intelligence between those who identify as
Sample sizes range from 1372 respondents in 1990 to 4510 Republican and those who identify as Democrat for three definitions of party
identity.
respondents in 2006; the mean sample size is just under 2000
respondents. Narrow Intermediate Broad
definition definition definition
2.2. Measures Without covariates 5.48⁎⁎⁎ 3.47⁎⁎⁎ 2.47⁎⁎⁎
With covariates 1.26⁎⁎ 0.52⁎ −0.00
The primary measure of intelligence available in the GSS Observations 5985 14,887 20,025
is a 10-word vocabulary test in which the respondent is Notes: Each value is the Republican advantage in IQ points. Estimates are
asked to identify which of five phrases supplies the correct from weighted OLS models of verbal intelligence Covariates: age, age squared,
definition of a given word (see Smith, Marsden, Hout, & Kim, gender, race, language, marital status, education, log of real household income,
region effects, year effects.
2012). Notwithstanding its brevity, the test has a correlation ⁎ 5% Significance level, based on robust standard error.
of .71 with the Army General Classification Test (Wolfle, 1980). ⁎⁎ 1% Significance level, based on robust standard error.
In addition, there is a huge amount of psychometric evidence ⁎⁎⁎ 0.1% Significance level, based on robust standard error.
144 N. Carl / Intelligence 44 (2014) 142–148

Table 2 sexual relations between two adults of the same sex—do you
Difference in mean verbal intelligence between those who supported the think it is always wrong, almost always wrong, wrong only
Republican Party in elections and those who supported the Democratic Party
for voters and non-voters.
sometimes, or not wrong at all?” The second is assessed with
the question, “Do you think the use of marijuana should be
Voters Non-voters made legal or not?” The third is assessed with the question,
Without covariates 1.82⁎⁎⁎ 1.86⁎⁎⁎ “Tell me whether you think it should be possible for a pregnant
With covariates −0.42 0.34 woman to obtain a legal abortion if she wants it for any reason.”
Observations 14,231 5660 The fourth is assessed with the question, “Suppose an admitted
Notes: Each value is the Republican advantage in IQ points. Estimates are communist wanted to make a speech in your community.
from weighted OLS models of verbal intelligence. Covariates: age, age Should he be allowed to or not?” The fifth is assessed with the
squared, gender, race, language, marital status, education, log of real question, “If a person wanted to make a speech in your
household income, region effects, year effects.
⁎⁎⁎ 0.1% significance level, based on robust standard error. community claiming that blacks are inferior, should he be
allowed to or not?” And the sixth is assessed with the question,
“Consider a person who advocates doing away with elections
and letting the military run the country. If such a person
wanted to make a speech in your community, should he be
and those who did not vote, respectively. The first takes the allowed to or not?”
value ‘1’ if a respondent voted Republican in the most recent I utilise six measures of economically liberal beliefs,
election, and takes the value ‘0’ if she voted Democrat. The namely: attitude toward government provision of jobs, attitude
second takes the value ‘1’ if a respondent would have voted toward government assistance of industry, attitude toward
Republican in the most recent election, and takes the value government redistribution of income, attitude toward price
‘0’ if she would have voted Democrat. In each case, every controls, attitude toward labour unions, and attitude toward
respondent who was asked about more than one election military spending (Smith et al., 2012). The first is assessed with
retains the observation corresponding to the most recent the question, “Do you think it should or should not be the
election. government's responsibility to provide a job for everyone who
A number of covariates are included in some of the models, wants one?” The second is assessed with the question, “Do you
namely: age, age squared, gender, race, language, marital status, think it should or should not be the government's responsibil-
education, and household income, as well as region and year ity to provide industry with the help it needs to grow?” The
effects (Smith et al., 2012). The GSS distinguishes between three third is assessed with the question, “Do you think it should or
racial categories: “white”, “black” and “other”. It distinguishes should not be the government's responsibility to reduce
between five levels of educational attainment: “less than high income differences between rich and poor?” The fourth is
school”, “high school”, “junior college”, “bachelor” and “gradu- assessed with the question, “Do you think it should or should
ate”. It distinguishes between five marital statuses: “mar- not be the government's responsibility to keep prices under
ried”, “widowed”, “divorced”, “separated” and “never married”. control?” The fifth is assessed with the question, “To what
Household income is the natural log of a respondent's household extent do you agree or disagree that unions in this country
income, given in constant 2000 dollars. Beginning in 2006, the have too little power?” And the sixth is assessed with the
GSS began to sample Spanish speakers (n = 513), alongside question, “Are we spending too little, about the right amount,
English speakers. Language is a dummy variable equal to 1 if the or too much on the military, armaments and defence?”
interview was conducted in Spanish. Descriptive statistics on the measures of social and economic
I utilise six measures of socially liberal beliefs, namely: beliefs are provided in Appendix A.
attitude toward homosexuality, attitude toward marijuana
legalisation, attitude toward abortion, attitude toward free 2.3. Models
speech for communists, attitude toward free speech for racists,
and attitude toward free speech for militarists (Smith et al., In the first stage of the analysis, OLS regression is used to
2012). The first is assessed with the question, “What about compute the difference in mean verbal intelligence between

Table 3
Standardized effects of verbal intelligence on six measures of social beliefs.

Homosexual relations Marijuana should Women should Communists should Racists should not Militarists should not
are wrong be illegal not be able to not be able to speak be able to speak be able to speak freely
get an abortion freely freely

Without covariates −.239⁎⁎⁎ −.115⁎⁎⁎ −.168⁎⁎⁎ −.317⁎⁎⁎ −.208⁎⁎⁎ −.254⁎⁎⁎


With covariates −.165⁎⁎⁎ −.118⁎⁎⁎ −.092⁎⁎⁎ −.234⁎⁎⁎ −.139⁎⁎⁎ −.170⁎⁎⁎
Observations 13,610 12,610 12,589 14,024 12,783 12,774

Notes: Each value is the standardized coefficient on verbal intelligence. Estimates are from weighted OLS models. Covariates: age, age squared, gender, race,
language, marital status, education, log of real household income, region effects, year effects.
⁎⁎⁎ 0.1% significance level, based on robust standard error.
N. Carl / Intelligence 44 (2014) 142–148 145

Table 4
Standardized effects of verbal intelligence on six measures of economic beliefs.

Government Government Government Government Unions do not Military


should provide should assist should reduce should control have enough spending is not
a job for everyone industrial growth income differences prices power high enough

Without covariates −.267⁎⁎⁎ −.162⁎⁎⁎ −.219⁎⁎⁎ −.309⁎⁎⁎ −.273⁎⁎⁎ −.103⁎⁎⁎


With covariates −.154⁎⁎⁎ −.088⁎ −.131⁎⁎⁎ −.173⁎⁎⁎ −.210⁎⁎⁎ −.096⁎⁎⁎
Observations 2904 1342 2096 1375 807 13,409

Notes: Each value is the standardized coefficient on verbal intelligence. Estimates are from weighted OLS models. Covariates: age, age squared, gender, race,
language, marital status, education, log of real household income, region effects, year effects.
⁎ 5% significance level, based on robust standard error.
⁎⁎⁎ 0.1% significance level, based on robust standard error.

those who identify as Republican and those who identify as results imply that the difference in verbal intelligence between
Democrat, separately for the three definitions of party identity. Republicans and Democrats is largest when the comparison is
Differences are computed both before and after adjusting for restricted to those with the strongest partisan identity, and is
socio-economic characteristics. In the second stage of the smallest when the comparison is extended to those with the
analysis, OLS regression is used to compute the difference in weakest partisan identity. In addition, they suggest that
mean verbal intelligence between those who supported the the effect of verbal intelligence on party identity is largely
Republican Party in elections and those who supported the accounted for by socio-economic characteristics. People with
Democratic Party, separately for voters and non-voters. Again, higher verbal intelligence tend to have better education, higher
differences are computed both before and after adjusting for incomes, and are more likely to be married (Herrnstein &
socio-economic characteristics. In the third stage of the Murray, 1994; Deary, 2012). And such people are more likely to
analysis, verbal intelligence is included as an independent identify as Republican.
variable in OLS models of social and economic beliefs. Table 2 displays the difference in mean verbal intelli-
A relatively small number of respondents are intention- gence between those who supported the Republican Party in
ally excluded from the analysis. In 1982 and 1987, blacks elections and those who supported the Democratic Party,
were oversampled as part of a National Science Foundation separately for voters and non-voters. In the case of voters,
research project (Smith et al., 2012). Because the samples mean verbal intelligence is 1.82 IQ points (d = 0.12) higher
from these years are not representative of the U.S. popula- among Republicans. This difference reverses and ceases to be
tion, all oversampled respondents (n = 707) are excluded statistically significant when covariates are included. In the
from the analysis. Respondents for whom at least one of the case of non-voters, mean verbal intelligence is 1.86 IQ points
covariates took a missing value are also excluded. Estimates (d = 0.12) higher among Republicans. Again, this difference
from the models without covariates are qualitatively identi- ceases to be statistically significant when covariates are
cal if these respondents are included (results not shown). In included. The preceding results indicate that the effect of
addition, the regression models are weighted as a way of verbal intelligence on voting decision is entirely accounted for
compensating for bias due to unequal household size, and by socio-economic characteristics. As a robustness check,
bias due to non-respondent sub-sampling in 2004 and 2006 Appendix B repeats the analysis using only white respondents.
(Stephenson, 1978; Smith et al., 2012). The estimates are Table 3 displays standardized effects of verbal intelligence
qualitatively identical if unweighted models are estimated on social beliefs. There is strong evidence that Americans
(results not shown). with higher verbal intelligence tend to be more socially

3. Results

Table 1 displays the difference in mean verbal intelligence Table 5


Standardized effects of verbal intelligence on principal components of social
between those who identify as Republican and those who and economic beliefs.
identity as Democrat, separately for the three definitions of
party identity. Under the narrow definition, mean verbal Social conservatism Economic statism
intelligence is 5.48 IQ points (d = 0.37) higher among Without covariates −.350⁎⁎⁎ −.338⁎⁎⁎
Republicans. This difference falls to 1.26 IQ points (d = 0.08) With covariates −.264⁎⁎⁎ −.210⁎⁎⁎
when covariates are included. Under the intermediate defini- Observations 3220 1247
tion, mean verbal intelligence is 3.47 IQ points (d = 0.23) Notes: Each value is the standardized coefficient on verbal intelligence.
higher among Republicans. This difference falls to 0.52 IQ Estimates are from weighted OLS models. Social conservatism had an
points (d = 0.03) when covariates are included. Under the eigenvalue of 2.51 and explained 42% of the variance. Economic statism had
an eigenvalue of 2.25 and explained 56% of the variance. Covariates: age, age
broad definition, mean verbal intelligence is 2.47 IQ points squared, gender, race, language, marital status, education, log of real
(d = 0.16) higher among Republicans. This difference disap- household income, region effects, year effects.
pears completely when covariates are included. The preceding ⁎⁎⁎ 0.1% significance level, based on robust standard error.
146 N. Carl / Intelligence 44 (2014) 142–148

liberal. They are less likely to believe that homosexual al., 2008a, 2008b; Stankov, 2009; Kanazawa, 2010; Schoon et
relations are wrong (β = − .24), that marijuana should be al., 2010; Heaven et al., 2011; Hodson & Busseri, 2012) and
illegal (β = − .12), that women should not be able to get an less religious (Bell, 2002; Lynn, Harvey, & Nyborg, 2009;
abortion (β = − .17), that communists should not be able to Nyborg, 2009; Ganzach et al., 2013; Zuckerman et al., 2013).
speak freely (β = − .32), that racists should not be able to This should lead one to expect that Republicans are less
speak freely (β = − .21), and that militarists should not be intelligent than Democrats. However, looking at data from
able to speak freely (β = − .25). Table 4 displays standard- the General Social Survey, I find that Republicans have
ized effects of verbal intelligence on economic beliefs. Once slightly higher verbal intelligence than Democrats. In partic-
again, there is strong evidence that Americans with higher ular, individuals who identify as Republican have slightly
verbal intelligence tend to be more economically liberal. They higher verbal intelligence than those who identify as
are less likely to believe that the government should provide Democrat (2–5 IQ points), and individuals who supported
a job for everyone (β = − .27), that the government should the Republican Party in elections have slightly higher verbal
assist industrial growth (β = − .16), that the government intelligence than those who supported the Democratic Party
should reduce income differences (β = − .22), that the (2 IQ points). I reconcile these findings with the previous
government should control prices (β = − .31), that unions literature by showing that, consistent with Pinker's (2011)
do not have enough power (β = − .27), and that military hypothesis, Americans with higher verbal intelligence tend to
spending is not high enough (β = − .10). have more socially liberal beliefs and more economically
The fact that the correlation between verbal intelligence and liberal beliefs (β = .10–.32).
economically liberal beliefs persists after controlling for Overall, my findings suggest that higher intelligence
characteristics like race, education and income suggests it among classically liberal Republicans compensates for lower
cannot simply be attributed to selfishness on the parts of people intelligence among socially conservative Republicans. Inter-
with higher verbal intelligence. In particular, it contradicts the estingly, the difference in verbal intelligence between those
hypothesis that such people only have economically liberal who supported the Republican Party in elections and those
beliefs because they believe they have personally benefitted who supported the Democratic Party (2 IQ points) is
from economically liberal policies. On the other hand, if somewhat smaller than the difference between those who
conditional on current income, verbal intelligence is correlated identify as Republican and those who identify as Democrat
with a tendency to believe one's future income will be higher (2–5 IQ points). One possible explanation is that Indepen-
under economically liberal policies, the correlation between dents with higher intelligence are more likely to support
verbal intelligence and economically liberal beliefs could be the Democratic Party in elections. Future research should
explained by selfishness, at least in part. test this prediction. It should also address the more com-
Table 5 displays standardized effects of verbal intelligence plex question of why intelligence happens to be correlated
on a principal component of social beliefs (social conservatism) with particular kinds of social and economic beliefs
and a principal component of economic beliefs (economic (Charlton, 2009; Kanazawa, 2010; Woodley, 2010; Dutton,
statism). Social conservatism was obtained by extracting the 2013).
first principal component from a PCA on all six measures of Two important caveats should be attached to my
social beliefs. Economic statism was obtained by extracting the conclusions. First, a 10-word vocabulary test is at best an
first principal component from a PCA on four measures of imperfect measure of verbal intelligence, let alone general
economic beliefs: attitude toward government provision of intelligence. Indeed, there is already evidence that the
jobs, attitude toward government assistance of industrial correlation between intelligence and political beliefs varies
growth, attitude toward government redistribution of income, with the sub-dimension of intelligence under scrutiny.
and attitude toward price controls. The two other measures of Specifically, Kemmelmeier (2008) found that although
economic beliefs were omitted because including them libertarian social attitudes were positively related to verbal
dramatically reduced the number of available cases. Insofar as SAT scores, they were unrelated to math SAT scores. Second,
many of the questions on social and economic beliefs were most of the effects I observe are small by conventional
asked at different waves, it was not possible to estimate models standards, especially the differences in verbal intelligence
using a single dimension of classically liberal beliefs. Verbal between Republicans and Democrats. According to conven-
intelligence has a moderate negative effect on both social tion, a standardized difference of 0.20 constitutes a small
conservatism and economic statism. Before adjusting for effect size; by contrast, a standardized difference of 0.50 is
covariates, β = −.35 in the model of social conservatism and considered moderate, while a standardized difference of 0.80
β = −.34 in the model of economic statism. After adjusting for or greater is considered large (Cohen, 1988, pp. 24–8).
covariates, β = −.26 in the model of social conservatism and
β = −.21 in the model of economic statism. For further
discussion of the relationships between intelligence and social Acknowledgements
beliefs, and between intelligence and economic beliefs, I refer
the reader to Kanazawa (2010) and Caplan and Miller (2010), I acknowledge support from the University of Oxford,
respectively. from Nuffield College, Oxford, and from the Economic and
Social Research Council (UK). I am grateful to the National
4. Discussion Opinion Research Centre for making their data available to
researchers. I would like to thank Bryan Pesta and two
Previous studies have documented that people with anonymous reviewers for providing comments on a previous
higher intelligence tend to be more socially liberal (Deary et version of the manuscript.
N. Carl / Intelligence 44 (2014) 142–148 147

Appendix A

Table A.1
Descriptive statistics for measures of social and economic beliefs.

Minimum Maximum Mean Standard deviation Observations

Homosexual relations are wrong 1 4 3.14 1.26 13,610


Marijuana should be illegal 1 2 1.72 0.45 12,610
Women should not be able to get an abortion 1 2 1.59 0.49 12,589
Communists should not be able to speak freely 1 2 1.35 0.48 14,024
Racists should not be able to speak freely 1 2 1.38 0.48 12,783
Militarists should not be able to speak freely 1 2 1.37 0.48 12,774
Government should provide a job for everyone 1 4 2.34 1.03 2904
Government should assist industrial growth 1 4 2.78 0.79 1342
Government should reduce income differences 1 4 2.38 1.06 2096
Government should control prices 1 4 2.87 0.87 1375
Unions do not have enough power 1 4 2.24 0.77 807
Military spending is not high enough 1 3 1.88 0.74 13,409

Appendix B Table B.2 displays the difference in mean verbal intelli-


gence between whites who supported the Republican Party
It is well-documented that blacks score lower than whites on in elections and whites who supported the Democratic Party,
the GSS measure of verbal intelligence (Lynn, 1998; Huang & separately for voters and non-voters. In the case of voters,
Hauser, 2001). Part of this disparity is due to downward bias in the Republican advantage is negative, while in the case of
the test scores of blacks (Huang, 2009). There may be other racial non-voters, it is very small. This suggests that the difference
biases in the GSS measure of verbal intelligence as well. To check in verbal intelligence by voting decision may be entirely
whether the results concerning Republicans and Democrats are attributable to lower test scores among non-whites. Overall,
attributable to lower test scores among non-whites, who are the results from Tables B.1 and B.2 reinforce the prediction
more likely to be Democrats (Newport, 2013), I repeat the that Independents with higher intelligence are more likely to
analysis using only white respondents. support the Democratic Party in elections.
Table B.1 displays the difference in mean verbal intelli-
gence between whites who identify as Republican and whites
who identity as Democrat, separately for the three definitions Table B.2
Difference in mean verbal intelligence between whites who supported the
of party identity. Under the narrow definition, mean verbal
Republican Party in elections and whites who supported the Democratic
intelligence is 3.32 IQ points (d = 0.22) higher among Repub- Party for voters and non-voters.
licans before adjustments, and 1.6 IQ points (d = 0.11) higher
after. Under the intermediate definition, mean verbal intelli- Voters Non-voters
gence is 2.01 IQ points (d = 0.13) higher among Republicans Without covariates −0.25 0.73
before adjustments, and 0.71 IQ points (d = 0.05) higher after. With covariates −0.43 1.00⁎
Under the broad definition, mean verbal intelligence is 1.2 IQ Observations 12,174 4324

points (d = 0.08) higher among Republicans before adjust- Notes: Each value is the Republican advantage in IQ points. Estimates are
ments, and not significantly higher after. The preceding esti- from weighted OLS models of verbal intelligence. Covariates: age, age
squared, gender, language, marital status, education, log of real household
mates are smaller than those in Table 1, which suggests that
income, region effects, year effects.
the difference in verbal intelligence by party identity may ⁎ 5% significance level, based on robust standard error.
be partly attributable to lower test scores among non-whites.
However, a statistically significant Republican advantage re-
mains even when non-white respondents are excluded.
References

Bell, P. (2002). Would you believe it? Mensa Magasine, 12–13 (Feb.).
Table B.1 Berezow, A., & Campbell, H. (2012). Science left behind: Feel-good fallacies and
Difference in mean verbal intelligence between whites who identify as the rise of the anti-scientific left. Philadelphia: Public Affairs, Perseus
Republican and whites who identify as Democrat for three definitions of Books Group.
party identity. Berlin, I. (1969). Four essays on liberty. Oxford: Oxford University Press.
Caplan, B. (2001). What makes people think like economists? Evidence on
Narrow Intermediate Broad economic cognition from the “Survey of Americans and Economists on
definition definition definition the Economy”. Journal of Law and Economics, 44, 395–426.
Caplan, B. (2007). The myth of the rational voter. Princeton: Princeton
Without covariates 3.32⁎⁎⁎ 2.01⁎⁎⁎ 1.20⁎⁎⁎ University Press.
With covariates 1.60⁎⁎⁎ 0.71⁎⁎ 0.24 Caplan, B., & Miller, S. C. (2010). Intelligence makes people think like
Observations 4600 12,192 16,667 economists: Evidence from the General Social Survey. Intelligence, 38,
636–647.
Notes: Each value is the Republican advantage in IQ points. Estimates are Caplan, B., & Miller, S. C. (2012). Positive versus normative economics:
from weighted OLS models of verbal intelligence. Covariates: age, age What's the connection? Evidence from the Survey of Americans and
squared, gender, language, marital status, education, log of real household Economists on the Economy and the General Social Survey. Public Choice,
income, region effects, year effects. 150, 241–261.
⁎⁎ 1% significance level, based on robust standard error. Carl, N., & Billari, F. C. (2014). Generalized trust and intelligence in the
⁎⁎⁎ 0.1% significance level, based on robust standard error. United States. PLoS One, 9, e91786.
148 N. Carl / Intelligence 44 (2014) 142–148

Cattell, R. B. (1963). Theory of fluid and verbal intelligence: A critical Kohut, A., Doherty, C., & Dimmock, M. (2009). Scientific achievements less
experiment. Journal of Educational Psychology, 54, 1–22. prominent than a decade ago. Pew Research Centre for the People & the
Charlton, B. G. (2009). Clever sillies: Why high IQ people tend to be deficient Press (July 9).
in common sense. Medical Hypotheses, 73, 867–870. Kohut, A., Doherty, C., Dimmock, M., & Keeter, S. (2012). What the public
Cohen, J. (1988). Statistical power analysis for the behavioural sciences. knows about the political parties. Pew Research Centre for the People &
Hillsdale: Lawrence Erlbaum Associates, Inc., Publishers. the Press (April 11).
Deary, I. J. (2012). Intelligence. Annual Review of Psychology, 63, 453–482. Lynn, R. (1998). Has the black–white intelligence differences in the United
Deary, I. J., Batty, G. D., & Gale, C. R. (2008a). Bright children become States been narrowing over time? Personality and Individual Differences,
enlightened adults. Psychological Science, 19, 1–6. 25, 999–1002.
Deary, I. J., Batty, G. D., & Gale, C. R. (2008b). Childhood intelligence predicts Lynn, R., Harvey, J., & Nyborg, H. (2009). Average intelligence predicts
voter turnout, voting preferences, and political involvement in adult- atheism rates 137 nations. Intelligence, 37, 11–15.
hood: The 1970 British Cohort Study. Intelligence, 36, 548–555. McLean, I., & McMillan, A. (2009). The Concise Oxford Dictionary of politics.
Dutton, E. (2013). The cultural mediation hypothesis: A critical examination. Oxford: Oxford University Press.
Intelligence, 41, 321–327. Miller, D. (2003). Political philosophy: A very short introduction. Oxford:
Friedman, M. (1962). Capitalism and freedom. Chicago: University of Chicago Oxford University Press.
Press. Mooney, C. (2005). The Republican war on science. Cambridge: Basic Books,
Ganzach, Y., Ellis, S., & Gotlibovski, C. (2013). On intelligence, education and Perseus Books Group.
religious beliefs. Intelligence, 41, 121–128. Mooney, C. (2012). The Republican Brain: The science of why they deny science—
Heaven, P. C. L., Ciarrochi, J., & Leeson, P. (2011). Cognitive ability, right-wing and reality. Hoboken: John Wiley & Sons, Inc.
authoritarianism, and social dominance orientation: A five-year longi- Newport, F. (2007). An abiding relationship. Gallup: Republicans and religion
tudinal study amongst adolescents. Intelligence, 39, 15–21. (June 14).
Herrnstein, R. J., & Murray, C. (1994). The Bell Curve: Intelligence and class Newport, F. (2013). Democrats racially diverse. Gallup: Republicans mostly
structure in American life. New York: Free Press. white (February 8).
Hodson, G., & Busseri, M. A. (2012). Bright minds and dark attitudes: Lower Nyborg, H. (2009). The intelligence–religiosity nexus: A representative
cognitive ability predicts greater prejudice through right-wing sample of white adolescent Americans. Intelligence, 37, 81–93.
ideology and low inter-group contact. Psychological Science, 23, Pinker, S. (2011). The better angels of our nature. London: Allen Lane, Penguin
187–195. Books Ltd.
Horn, J. L., & Cattell, R. B. (1966). Refinement and test of the theory of fluid Rindermann, H., Flores-Mendoza, C., & Woodley, M. A. (2012). Political
and crystallized general intelligences. Journal of Educational Psychology, orientations, intelligence and education. Intelligence, 40, 217–225.
57, 253–270. Saad, L. (2012). Conservatives remain the largest ideological group in U.S.
Huang, M. (2009). Race of the interviewer and the black–white test score Gallup. (January 12).
gap. Social Science Research, 38, 29–38. Schoon, I., Cheng, H., Gale, C. R., Batty, G. D., & Deary, I. J. (2010). Social status,
Huang, M., & Hauser, R. M. (2001). Convergent trends in black–white verbal cognitive ability, and educational attainment as predictors of liberal
test-score differentials in the U.S.: Period and Cohort Perspectives. social attitudes and political trust. Intelligence, 38, 144–150.
EurAmerica, 31, 185–230. Smith, T., Marsden, P. V., Hout, M., & Kim, J. (2012). General Social Surveys,
Iyer, R., Koleva, S., Graham, J., Ditto, P., & Haidt, J. (2012). Understanding 1972–2012 Cumulative Codebook. Chicago: National Opinion Research Centre.
libertarian morality: The psychological dispositions of self-identified Stankov, L. (2009). Conservatism and cognitive ability. Intelligence, 37, 294–304.
libertarians. PLoS One, 7, e42366. Stephenson, B. (1978). Weighting the General Social Surveys for bias related
Jensen, A. (2001). Vocabulary and general intelligence. Behaviour and Brain to household size. GSS methodological report no. 3. Chicago: National
Sciences, 24, 1109–1110. Opinion Research Centre.
Kanazawa, S. (2010). Why liberals and atheists are more intelligent. Social Wolfle, L. (1980). The enduring effects of education on verbal skills. Sociology
Psychology Quarterly, 73, 33–57. of Education, 53, 104–114.
Kemmelmeier, M. (2008). Is there a relationship between political orientation Woodley, M. A. (2010). Are high-IQ individuals deficient in common sense? A
and cognitive ability? A test of three hypotheses in two studies. Personality critical examination of the ‘clever sillies’ hypothesis. Intelligence, 38, 471–480.
and Individual Differences, 45, 767–772. Zuckerman, M., Silberman, J., & Hall, J. A. (2013). The relation between
Kirby, D., & Boaz, D. (2010). The libertarian vote in the age of Obama. Cato intelligence and religiosity: A meta-analysis and some proposed
Institute Policy Analysis, 658, 1–19. explanations. Personality and Psychology Review, 17, 325–354.

Vous aimerez peut-être aussi